Global rule induction for information extraction

Abstract

Information Extraction (IE) is designed to extract specific data from high volumes of text, using robust means. Pattern rule induction is one kind of techniques which have been widely used in IE. This thesis focuses on pattern rule induction for IE on both semi-structured and free texts. First, we introduce GRID, a Global Rule Induction for text Documents, which emphasizes on utilizing the global feature distribution of all of the training examples to start the rule induction process. Then, we show GRID can be applied successfully in definitional question answering and video story segmentation tasks. Lastly, we introduce two weakly supervised learning paradigms by using GRID as the base learner. One weakly supervised learning scheme is realized by combing co-training GRID with two views and active learning. The other weakly supervised learning paradigm is implemented by cascading use of a soft pattern learner and GRID.